Multimodal image re-registration via mutual information to account for initial tissue motion during image-guided neurosurgery

Multimodal imaging modalities, such as high-resolution pre-operative MR (pMR) and intra-operative US (iUS), have been used widely to provide complementary information for guiding neurosurgery. Proper registration between the images is usually established fiducials placed on the patient skull, which can be identified both in MR and a tracking system that also records location and orientation of each iUS scan. Although such systems can achieve high registration accuracy, we have observed brain tissue motion with respect to the skull after the craniotomy and before opening of the dura. We developed a rigid registration strategy to re-align the brain in iUS and pMR to account for such motion. The mutual information between the two modalities is maximized through a genetic algorithm. Two patient cases are presented to show the improved registration accuracy of soft tissues.